Finite and non-finite complementation, particles and control in Aromanian,
compared to other Romance varieties and Albanian
Our Aromanian data come from Diviakë, Libofshë and Fier, three locations close to one another in
South Albania, and from Këllez, also in South Albania. Section 1, on finite complementizers, shows
that Aromanian supports analyses of complementizers as wh- pronouns independently developed
for other Romance languages. Section 2 elaborates a proposal originally put forth for Albanian,
where the subjunctive particle is identical to the Linker – namely that the main role of subjunctive
particles is introducing a variable EPP argument, subject to control. Section 3 argues that the
impossibility of embedding sentences directly under V-v (a sort of ‘Agree Resistance Principle’)
leads to the overall shape of complementation in Romance, determining the obliquization strategy
represented by Prepositional introducers of non-finite sentences – as well as to the nominalization
(relativization) strategy of section 1.
1.
Finite complementation
1.1
Complementizers as wh- pronouns
All four Aromanian varieties under consideration, Diviakë (D), Libofshë (L), Fier (F) and Këllez
(K), have a finite complementizer kə which has the same general properties as standard Romanian
că /kə/, introducing indicative complements to verbs of knowing, saying etc., as illustrated in (1).
(1)
a.
a’.
b.
c.
c’.
d.
mini mundɛsk kə atseu u zinə məni
I
think that he
will come tomorrow
‘I think that he will come tomorrow’
mini esku kutʃunit kə
atseu (nu) (ma)
dɔrm
I
am
happy that he
not Progr
sleeps
‘I am happy that he is not sleeping’
m
aspari
kə
vini
asɛra
to.me it.seems
that he.came
yesterday
‘It seems to me that he came yesterday’
ɲ
ar
dzəs kə
eu
vini
to.me they.have
said that he
came
‘They told me that he came’
nu
o
ʃtiu
kə
tsi
məki eu
not
it
I.know that what eats he
‘I don’t know what he eats’
am vədzut
kə
eɫ
ari
faptə mult luk
I.have seen that he
has
done a lot of work
‘I saw that he did a lot of work’
D
L
F
K
A second finite complementizer of standard Romanian, namely ca, co-occurs with the
subjunctive particle să; ca să may be adjacent in Early Romanian (Hill 2013a), though in standard
Romanian ca and să must be separated by Topic/Focus material. In the varieties of Diviakë,
Libofshë and Fier, the second finite complementizer, with essentially the same distribution as
Romanian ca, takes the form tsi, as in (2)-(4). Normally it co-occurs with the subjunctive particle
si/sə. The latter is independently known to be compatible with control and non-control readings of
the embedded sentence – and this remains true when the tsi complementizer is present. For
instance, control examples with ‘want’ include (4a), while non-control readings with the same verb
include (2a)-(3a). In most examples tsi and the subjunctive particle are directly adjacent, though
they can also be separated by a Topic or Focus, for instance in (2a), or (3a-b).
(2)
a.
b.
c.
d.
e.
(3)
a.
b.
c.
d.
e.
e’.
(4)
a.
b.
c.
mini vɔi
tsi
atseu s
(nu) u
məkə
I
want that he
Prt
not
it
eats
‘I want him to eat it’
mini ʎi
dzɛs tsi
sə
ɣa
straɲəʎi
I
to.him say
that Prt
he.wash
the.clothes
‘I tell him to wash the clothes’
am
agrəʃətə
tsi
s
mək
I.have forgotten
that Prt
I.eat
‘I forgot to eat’
am
inʃɛtə
ka ma ninti tsi
s
mək
I.have gone.out
more before that Prt
I.eat
‘I left before eating’
esti pənə
tsi
s
mə"kəm
is bread
that Prt
we.eat
‘It is bread for us to eat’
vɔi
tsi
tini
z
iɲ
məni
I.want that you Prt
come tomorrow
‘I want you to come tomorrow’
ma
tsə
dzɛk atsea tsi
s
o
bei
Progr to.you I.say to.you that Prt
it
you.drink
‘I tell you to drink it’
o cisiə
tsi
s
o
fak
mini
it I.started
that Prt
it
I.make I
‘I began to do it’
am
vənit tsi
s
ti vəd
I.have come that Prtyou I.see
‘I have come to see you’
esti
pəni tsi
nu
s
məki
it.is bread that not
Prt
he.eats
‘It’s not bread for him to eat’
o
fɛtʃ
tʃələmənu
tsi
z
inə
ɣa mini
him I.made
the.boy
that Prt
comes to me
‘I made the boy come to me’
mini vɔi
tsi
s
dɔrm (mini)
I
want that Prt
sleep I
‘I want to sleep’
eu
agərʃɛ (tsi) si
z
ɣa
kəmɛʃa
he
forgot that Prt
Prt
he.wash
the.shirt
‘He forgot to wash the shirt’
eu
vini dənint tsi
s
vənɛm
mini
he
came before that Prt
came
I
‘He came before I did’
D
L
F
Note however that tsi need not co-occur with the subjunctive particle, as illustrated in (5).
(5)
a.
ɲ
ar
to.me they.have
dzəs
said
tsi
that
eu
he
vini
came
F
b.
c.
‘They told me that he came’
ɲ
ar
dzɛs tsi
eu
ini
məni
to.me they.have
said that he
comes tomorrow
‘They told me that he comes tomorrow’
u
fəts tsi
skəpi
kənəli
him I.made that escapes
the.dog
‘I made the dog escape’
L
D
A fact which will be of particular interest for our analysis is that in Aromanian, tsi is also the
wh-pronoun for ‘what’, though specialized interrogatives like kai ‘who’, o kui ‘to whom’ are
equally present. As a relative pronoun, furthermore, tsi takes both animate and inanimate
antecedents (cf. French que, inanimate as an interrogative, but taking both animate and inanimate
restrictors as a relative). We illustrate these conditions with the data from Fier and Libofshë in (6)(7). Këllez, to whose complementizer system we will return, has similar data for wh-pronouns, as in
(8). An argument in favour of the conclusion that pronoun tsi and complementizer tsi are not merely
homophonous, but instantiate the same lexical item, comes from the fact that tsi cannot embed any
wh-item (including non-homophonous ones), though the sequence k«+wh- is attested, for instance
in (1c’) above.
(6)
a.
b.
c.
d.
e.
(7)
a.
b.
c.
(8)
a.
b.
c.
d.
tsi
fats
what you.do
‘What do you do?’
nu
ʃtiu
tsi
fak
not
I.know what I.do
‘I don’t know what I do’
nu
ʃtiu
tsi
s
fak
not
I.know what Prt
I.do
‘I don’t know what to do’
esti
atseu tsi
ɲə
grɛsti
he.is the.one that to.me he.calls
esti
atseu tsi
mini o
vəd
tətuna
he.is the.one that I
him see
always
tsi
fats
what you.do
‘What do you do?’
esti
eu
tsi
vədi tatuna
he.is the.one that I.see always
‘He is the one that I always see’
esti
eu
tsi
mi
ɣrɛʃti tatuna
he.is the.one that me
calls always
‘he is the one who always calls me’
tsi
fats
what you.do
‘What do you do?’
tsi
s
fak
what Prt
I.do
‘What should I do?’
ei
nu
ʃti
tsi
s
fake
they not
they.know
what Prt
they.do
‘They do not know what to do’
omu
tsi
mi
atʃamɛ totəna / tsi
ved
F
L
K
totəna
the man
that me
calls always/ that
‘The man that always calls me/that I always see’
I.see
always
In Këllez, the finite complementizer, occurring in the same environments as Romanian ca or
Aromanian tsi, takes the form ta, as in (9). This particle also surfaces in a few examples of Fier
where it precedes tsi (and s), as in (10).
(9)
a.
b.
c.
(10)
a.
b.
u
s
vram ta
z
vina
Fut
Prt
I.want Comp Prt
he.come
‘I would want him to come tomorrow’
i
dzɛk ta
s
zjinə
to.him I.say Comp Prt
he.come
‘I told him to come’
am
vinitə ta
s
ti
ved
I.have come Comp Prt
you I.see
‘I have come to see you’
esk
kutʃənit
(ta) tsi
ti
vɛd
I.am happy
Comp that you I.see
‘I am happy to see you’
mini iʃai
dininti (ta) tsi
z
I
went.out
before Comp that Prt
‘I left before you came’
mɛni
tomorrow
K
F
vənɛi
you.came
The coincidence between complementizer tsi and wh-pronoun tsi is to be evaluated against a
set of recent proposals that profoundly modify the conception of the category COMP/C. In the
classical generative categorization (Bresnan 1972), function is paramount, in the sense that the role
sentential introducers have in turning main sentences into complement sentences defines their
categorial content. In Kayne’s (1976) systematization of French, the lexical coincidence of
complementizer que and of relative pronoun que is taken into account – but resolved in favour of
treating both as COMP. In more recent years, the issue of the lexical coincidence of
complementizers and wh-pronouns in Romance has been re-examined, leading several scholars to
conclude that a complementizer is just (a special type of) a wh-pronoun. The observation that these
different elements are lexically identical is compatible with weaker explanations, including in
particular grammaticalization; Roberts and Roussou (2003) propose that this should be construed as
a process whereby the argument DP, i.e. the pronoun, shifts to functional category status (C, or one
of its cartographic articulations) (cf. also Bayer 2001).
The range of proposals we are interested in adopts the stronger view that complementation is
relativization – and therefore requires a wh-pronoun (a demonstrative in Germanic, etc.). Two
different approaches can be distinguished. The earlier approach of Manzini and Savoia (2003, 2005,
2011) takes wh- pronouns and complementizers to be lambda operators (the standard assumptions
for wh- pronouns). They are differentiated from one another in that the wh-pronouns introduces an
individual variable, while the complementizer introduces a propositional variable, restricted by the
embedded propositional content. Consider for instance the direct question of Fier in (6a) and the
complement sentence in (4a). In (6a), with the structure in (11), tsi introduces an individual variable
corresponding to the internal argument of the verb. The structural role of tsi is the same in (12),
corresponding to example (4a) – except that the variable has propositional content.
(11)
(12)
[tsi x [fats x]
[mini vɔi
[tsi x, x: s dɔrm mini]
A variant of the same general idea is proposed by Arsenjievic (2009), Kayne (2010),
according to whom sentential complements are relative clauses headed by an abstract noun. They
argue that Complex NPs such as the fact/ the claim /the belief etc. that should be analysed as
relative clauses. They then proceed to analyse all sentential complements as involving an abstract
version of these nominals. In essence, while Arsenjievic and Kayne propose that all complement
clauses are in reality relatives, Manzini and Savoia can be taken to argue that they are rather
headless (free) relatives.
Before proceeding we should acknowledge a number of issues that these treatments raise.
One set of issues relates to the status of the C category, once complementizers and wh-pronouns are
identified, hence obviously assigned to the same category, say Q(P), leaving the C category
potentially deprived of membership. In reality, Rizzi’s (1997) influential proposal on the
cartographic organization of the left periphery already categorizes one C position as Fin,
acknowledging its verb-related nature. This is the position targeted by V2 phenomena – while a
higher C position (perhaps the Rizzian ‘Force’) appears to be targeted by imperatives. Given that
these V-related positions are by now embedded into generative theory as Cs we may keep this label
for them.1 For similar reasons of descriptive ease, we keep the descriptive term complementizer
rather than substituting it with clumsier periphrases such ‘propositional relativizer’ etc.
Another issue to be considered is that both conventional relative clauses and free relatives
create islands – while finite complement sentences do not. Here Manzini and Savoia’s idea that
internally-headed free relatives are involved may turn out to be useful. We assume that the
embedded structure that we have arrived at in (12) is all that is required to build a propositional free
relative, hence a complement sentence. Consider however relatives or free relatives built on an
individual variable; we may want to assume that this requires an abstract DP layer of structure. It is
this additional layer that creates the DP-over-CP (bi-phasal) structure corresponding to the
(Complex NP) Island. One may also wonder about the status of complement clauses vis-à-vis whislands, assuming that the latter violate (Relativized) Minimality (Rizzi 1990, Chomsky 1995).
Again, as in the discussion that precedes, the content of the wh- pronoun may be relevant;
specifically, it is reasonable to propose that the propositional wh- operator that we conventionally
call a complementizer does not count as an intervener for wh- arguments and adjuncts.
It must also be mentioned that though the lexical overlapping of wh- pronouns and
complementizers in many Romance languages provided an important clue as to the real nature of
the latter, several Romance languages have finite complementizers that are not syncretic with whpronouns at all. In fact, standard Romanian is a case in point. Thus că and ca are distinct from the
wh-pronouns series care ‘which’, ce ‘what’, cine ‘who’, cui ‘to which’. We take it that the lack of
lexical coincidence simply means that a specialized lexicalization is available for the propositional
wh- pronoun (the so-called complementizers) and wh-pronouns introducing individual variables. As
we will see in the next section, another fact concerning specifically two-complementizer systems is
rather more interesting from a theoretical point of view, namely that only one of the two
complementizers overlaps with wh- pronouns.
1.2
1
Two-complementizer systems
A related issue is whether we can derive the complementary distribution of complementizers and V2 in German (den
Besten 1984) if C hosts V2, but not complementizers. A crucial observation is in many languages, V2 and overt
complementizers co-occur (e.g. Yiddish, Santorini 1989). An equally important fact is that though the absence of
complementizers may be a necessary condition for V2 in some languages, it is not a sufficient condition in any
language, V2 being restricted by syntactico-semantic constraints on ‘embedded root phenomena’ (Haegeman 2012).
These considerations potentially weaken the original argument in favour of complementizers and V2 being hosted by
the same C functional head.
A two-complementizer system similar to that of Aromanian is documented for the in-contact
language, Tosk Albanian (Manzini and Savoia 2007). In the variety of Gjirokaster, sentential
complements are introduced by sE and tS«; sE is excluded from most contexts including the
subjunctive particle t«, as in (13a), which require tS«, as in (13b). The latter overlaps with the whpronoun for ‘what’, as in (13d); this holds true despite the fact that standard orthography
distinguishes që (the complementizer) from ç ‘what’. In any event tS«/që also introduces relative
clauses, as in (13c).
(13)
a.
b.
c.
d.
jam ɛ
sigurt sɛ
dɔ
vijə
I.am fs.
sure that fut.
he.comes
‘I am sure that he will come’
tə
θatʃ tʃə
t
a
laɲɛ
to.you I.said that Prt
him/her
you.washed
‘I told you to wash him/her’
ai
ɐʃt
buri
tʃə
tə
θərrɛt
he
is
the man
that you he.calls
‘He is the man that always calls you’
tʃə
məndɔn
sɛ
tʃə
bən
ai
what you.think
that what do
he
‘What do you think that he is doing?’
ʃpɛʃ
always
Gjirokastër
Yet the similarity of Aromanian and Albanian does not appear to be a contact phenomenon,
or a Balkan feature, since similar conditions characterize Romance varieties of Central and
Southern Italy and Sardinia (Ledgeway 2003, 2005, Manzini and Savoia 2005, 2011). In Italian
varieties the complementizer occurring with complements to verbs of knowing, saying, believing
etc. like Romanian că or Aromanian kə, is normally ka, as in (14a), (15a). The complementizer
introducing complements of verbs of wanting, commanding, of before/after adjuncts etc., much like
Romanian ca and Aromanian tsi, is normally k« or ki in Italian varieties, as in (14b), (15b).
(14)
a.
b.
(15)
a.
b.
m
ann« D«tt« ka
u kES«
ts« l ann« t•t« i ÄwajjEun«
to.me they.have said that the cheese
MP it have taken the boys
‘They told me that the boys took the cheese’
vujj« k«
krE
pur« l•r«
m«n«ss«n«
I.want that tomorrow
also they
come
‘I want them as well to come tomorrow’
Guglionesi (Abruzzi)
m
anta
nau
ka
u libru
ˆaza
leddzju
to.me they.have
said that the book
it you.have read
‘They told me that the book you read’
EstE mmendzus
ki
fintsaza juanni
bEndzEDE
kraza
it.is better
that even John
comes
tomorrow
‘It is better that John as well comes tomorrow’
Paulilatino (Sardinia)
Ledgeway (2003, 2005), who refers to the two Southern Italian complementizers as CA and
CHE, finds that in a corpus of 11 early texts, “out of a total of 327 examples of CA-clauses … a
mere 10.1% were found to contain one or more elements in the left periphery, whilst from a total of
1061 examples of indicative clauses introduced by CHE, … 41,8% were found to host one or more
elements in the left periphery” (2005: 360). Adopting Rizzi’s (1997) model of the left periphery,
Ledgeway argues that “CA and CHE are invariably merged in … Fin° … as an overt reflex of the
different modal specifications (indicative vs subjunctive) they check in Fin°, from which
automatically follows the claim that CA only occurs in indicative clauses and CHE only in
subjunctive clauses’” (372). However, “the appearance of topics and foci forces the
complementizers CA and CHE to move out of Fin° to target’ Force°” (376); “this movement
operation is spelt out morphologically only in the case of the indicative complementizer CA which
invariably surfaces as CHE’ (374-375). In other words, for Ledgeway, the bipartition of
complementizers is defined by Rizzi’s categories of Fin and Force, so that CA only occurs in the
Fin position, while CHE occurs both in Fin and in Force.
In general, cartographic approaches are based on the idea that once morphemes and lexical
items are properly sequenced, there is only one possible mapping to the universal template of
functional categories. Therefore the correct sequencing of morphemes/lexical items reveals their
intrinsic properties as well, as encoded in the universal functional hierarchy. For Ledgeway, CA can
only be sequenced as Fin, since it occurs to the right of left peripheral material, while CHE can be
sequenced as Force. The sequencing effectively yields their content, in a classical demonstration of
cartographic reasoning. Manzini and Savoia (2011) however find that the Italian varieties they
consider do not display the asymmetries observed by Ledgeway, since foci and topics appear both
under ka and under k«/ki. This despite the fact that modern varieties otherwise match early varieties,
of which they are historical continuators. The examples in (14)-(15) already illustrate the relevant
facts, since both complementizers embed topics/foci. Therefore Manzini and Savoia conclude that
the two complementizers occupy very much the same position – essentially as in non-articulated
models of the left periphery with a single C slot (Chomsky 1995 ff.).
Insights on double complementizer systems can be gained from looking at their intrinsic
lexical content, rather than at their position. While the ka complementizer specializes in the role of
sentence introducer, the k«/ki complementizer overlaps with wh- pronouns in interrogatives and in
relatives, as in Abruzzese (16), or at least in relatives, as in Sardinian (17).
(16)
a.
b.
(17)
a.
b.
k«
ffi
what you.do
'What are you doing?'
E
kkull« k«
vvad« sEmbr«
he.is that that I.see always
‘He is the one that I see all the time’
Es
kussu ki
bbi• zEmpErE
he.is that that I.see always
‘He is the one that I see all the time’
kiE
bbeniDi
who comes
‘Who comes?’
Guglionesi
Paulilatino
Manzini and Savoia (2011) argue that “the ka complementizer is a definiteness element,
effectively the counterpart to a definite determiner”, so that the structure of the Guglionesi example
in (14a), with ka lexicalized, would be of the type in (18a). By contrast, the structure of the
Guglionesi example in (14b), with k@ lexicalized “could include an indefinite quantification,
corresponding to a free variable bound by existential closure”, as in (18b). In these terms, Manzini
and Savoia obtain “the generalization that of the two complementizers present in languages like
Guglionesi … it is always the indefinite complementizer that overlaps with a wh-quantifier. This is
because the other complementizer effectively introduces a definite description, which is hardly
compatible with wh-quantification – while the indefinite complementizer introduces a propositional
variable subject to existential closure, more or less like the argumental wh-variable”.
(18)
a.
b.
They told me the x: x the boys took the cheese
I want for some x: x they too come tomorrow
Vice versa, Manzini and Savoia observe that in at least some Italian varieties, there is no
one-to-one correlation between the lexicalization of the k@/ki complementizer and the subjunctive
inflection on the verb, since k@ can also embed an indicative, as in Abruzzese (19). In any event,
they argue that an indicative vs. subjunctive characterization of the two complementizers ka and k@
would not help in explaining why only the subjunctive complementizer would overlap with whpronouns, since the latter display no sensitive to the indicative/subjunctive distinction. Thus the
interrogative in (16a) is found in a main context which does not admit the subjunctive – and the
relative in (16b), (17a) is equally in the indicative.
(19)
pEnts@ k@
vE
krE
I.think that he.comes tomorrow
‘I think he comes tomorrow’
Guglionesi
The generalizations that Manzini and Savoia (2005, 2011) draw from Italian varieties, are
confirmed by Aromanian down to a considerable level of detail. First, there is evidence that
Topic/Focus material can follow both complementizers; thus k@ is followed by a wh-phrase in (1c’),
while sentences where tsi is followed by left peripheral material include (2a), (3a-b); at the same
time, left peripheral material is not necessary to the lexicalization of either complementizer. Second,
in some varieties (Fier, Libofshë, Diviakë) there is lexical coincidence between the complementizer
and the wh-pronouns system; if so, it is the complementizer that co-occurs with the subjunctive
particle tsi that coincides with a wh-pronoun. The importance of this generalization is that it makes
it unlikely that the coincidence of wh- pronouns and complementizers is an external accident;
rather, it appears to be rule-governed, and hence to warrant an internal explanation.
In recent years, the indicative/subjunctive distinction in Romance languages has been
clarified by the work of Giorgi (2009), according to whom it signals whether a representation of the
Speaker is present within the embedded sentence (indicative) or not (subjunctive), leading to
independent T anchoring in the indicative, but not in the subjunctive. In some languages, like
standard Italian or Sardinian, the morphological subjunctive encodes this distinction. However in
other languages like Abruzzese or Central/Southern Italian varieties the morphological subjunctive
appears to be restricted to irrealis readings proper, as in (14b). This however does not mean that the
relevant interpretive contrast between Speaker/T anchored contexts and non-anchored contexts is
not observed in this second set of languages – which means in turn that it could be relevant for the
choice of complementizers after all.
Suppose we keep to the conclusion that (12) represents the structure of a tsi sentence in
Aromanian. Based on Giorgi (2009), one may want to propose that Romance languages like
Aromanian (but also Abruzzese etc.) have a second complementizer, i.e. again a wh- operator,
which is restricted by sentential contents that are Tense- and Speaker- anchored (‘indicatives’). In
other words the alternation between k@ and tsi in Aromanian (ka and k@ in Abruzzese) etc. is rather
like the alternation between ‘who’ and ‘what’ in English – namely an alternation between two whoperators with different restrictions. To be more precise, it is normally accepted that English ‘who’
has a [human] restriction, while ‘what’ is unrestricted (answers to ‘what’ include animate and
inanimate individuals as well as events/situations). Similarly, for Aromanian, we can keep tsi as the
unrestricted complementizer and restrict k@, as just proposed, to Speaker/T anchored propositions.
A differentiation of two-complementizer systems according to their selection properties is
appealing for its simplicity. Morphological subjunctive properties do not appear to be involved –
since they seem to differ in their extension in Romance languages, while complementizer systems
are essentially stable. However abstract properties of Speaker/T anchoring – corresponding to
morphological indicative in a subset of languages – can adequately describe the selectional
restrictions of complementizers, as schematized for Aromanian in (20).
(20)
a.
b.
tsi:
k@:
whwh-, Speaker/T anchored proposition
We are now in a position to answer the question why lexical coincidence of
complementizers and wh- pronouns is always restricted to subjunctive complementizers. The
question turns out to be the converse to another one, namely why among wh- pronouns it is
generally ‘what’ that lexically overlaps with complementizer. The obvious answer is that ‘what’
lacking selectional restrictions is compatible with propositional content whereas ‘who’ is not.
Similarly a wh- element with the propositional restriction in (20b) could not turn up with an
individual variable interpretation. Recall lastly that in one of the Aromanian varieties we considered
namely Këllez, there is a complementizer k@ with the properties in (20b) – but no complementizer
tsi, while the distribution of tsi appears to be covered by a non-wh- element namely ta. Nevertheless
there is no reason why we couldn’t extend the wh- property to ta, and associate it with a
propositional restriction, that excludes it from introducing individual variables. We shall return to
the status of ta at the end of section 2.
1.3
Summary
Summarizing so far, Romance finite complementation involves turning sentences into free relatives,
headed by a wh- pronoun. The latter may be specialized for the task, but more often overlaps with
some wh- pronoun introducing individual variables – which provides the most obvious empirical
pointer to its nature. There are also Romance languages with two complementizers; we have
proposed in (20) that their distribution is governed by presence/absence of a restriction to Speaker/T
anchoring (Giorgi 2009). Only the unrestricted complementizer can overlap with a wh- pronoun
introducing an individual variable.
Several questions are raised by these proposals. The most important one is why finite
sentences in Romance would involve relativization. The answer seems to be the one originally
proposed by Rosenbaum (1967), namely that (finite, declarative) sentences are in fact not
embeddable and their nominalization is a strategy to circumvent this prohibition. Still one may want
to ask why exactly sentential contents are not embeddable – and how exactly the relativization
strategy used by finite embedding compares to other embedding strategies for instance in Romance.
We shall return to these question in section 3, after reviewing two further types of Aromanian
sentential complements, namely sentences introduced just by subjunctive particles and non-finite
(infinitival, participial) sentences.
2.
Subjunctive particles and control
2.1
The distribution of subjunctive particles
The process whereby Romanian progressively came to restrict the distribution of the (panRomance) infinitive in favour of clauses introduced by the subjunctive particle să is normally
interpreted as an instance of the pan-Balkan phenomenon of the loss of infinitive (Joseph 1983).
Note however that, as we will see in section 3, in standard Romanian infinitival complements are
present in the form of the short infinitive – and Aromanian has infinitival complements with–re
inflected infinitives (so-called long infinitives). In other words, there is no necessary connection
between Prt sentences and loss of infinitive.
Theoretical matters debated in connection with standard Romanian să concern notably its
status as a complementizer or as a Mood marker. For Dobrovie-Sorin (1994) să is generated in C,
thus accounting for its C-like characteristics, but I incorporates with C, thus accounting for the Ilike properties of să as well. Hill (2013a, 2013b) works in a different framework, including the split
CP analysis of Rizzi (1997), and has at her disposal two different C positions, namely Rizzi’s Fin
and Force. For her, the subjunctive particle să merges in the lower C position, i.e. Fin, while the ca
and că complementizers merge in Force. This means that Topic and Focus material precedes să and
follows ca/că. On the basis of the same data, Jordan (2009) reaches a different conclusion, namely
that să is an I/M element, following in essence Rivero’s (1994) proposal that subjunctive particles
occupy a dedicated projection MoodP.
Importantly, the presence of Prt makes the control reading of the embedded complement
possible. With obligatory control verbs (OC), să clauses display obligatory coreference of the
embedded subject with a matrix argument; să complements of non-obligatory control verbs, on the
other hand, can either display control or disjoint reference (Jordan 2009). Even when a non-control
reading is available, and a nominative subject can therefore be lexicalized, the latter can only appear
in peripheral position. Thus it can be postverbal or it can be topicalized between ca and să, but it
cannot intervene between the particle and the verb.
Aromanian examples where the s particle co-occurs with the tsi/ta complementizer have
already been provided in section 1.1. The presence of the particle allows the control reading –
though it is equally compatible with the presence of a disjoint subject. The two readings are
independent of the presence of tsi/ta, since analogous data can be found with sentences introduced
only by s. This latter configuration is illustrated in some detail in (21)-(24).
(21)
a.
b.
c.
d.
(22)
a.
b.
c.
d.
(23)
a.
b.
mini voj
z
in
I
want Prt
I.come
‘I want to come’
ts
am
dzəsə sə
(nu) ti
ɣajə
to.you I.have said Prt
(not) you you.wash
‘I told you to wash up’
esti
ɟini
s
u
vədəm
is
well Prt
it
we.see
‘It is good that we see it’
mini u
snekw z
dɔrm
I
Fut
I.go Prt
I.sleep
‘I will go to sleep’
nu
voi
s
ti
ved
not
I.want Prt
you I.see
‘I don’t want to see you’
ma
tsə
dzɛk s
o
fats
Progr to.you I.say Prt
it
you.make
‘I am telling you to do it’
ɲ
ɛrəseʃti
(tsi) s
o
ved
to.me it.likes
that Prt
him I.see
‘I like to see him’
mini fug
nu
s
vədə eu
I
go.away
not
Prt
I.see him
‘I go not to see him’
mini vrɛm
s
o
vədɛm
I
wanted
Prt
him I.saw
‘I wanted to see him’
mini mənduɛsk
s
u
vɛd
eu
I
think
Prt
him I.see him
‘I think I see him’
D
L
F
c.
d.
e.
(24)
a.
b.
c.
d.
e.
mini filli
(tsi) s
lədzɛsk
un libər
I
began that Prt
I.read
a book
‘I began reading a book’
eu
(nu) pɔt
z
in
he
not
can
Prt
he.comes
‘He can’t come’
nu
tə
ɣas
s
o
fats
not
you I.let Prt
it
you.do
‘I don’t let you do it’
vrɛu
(ta) s
u
məkə
I.wanted
that Prt
it
he.eats
‘I wanted him to eat it’
ts
dzɛʃ (ta) s
u
fats
to.you I.said that Prt
it
you.make
‘I told you to do it’
buna esti
s
durɲim
good is
Prt
we.sleep
‘It is good that we sleep/ to sleep’
nɛg (ta) s
u
vɛd
I.go that Prt
him I.see
‘I go to see him’
u
fets s
dɔʁmə
him I.made Prt
he.sleeps
‘I made him sleep’
K
The si particle occurs lower than the complementizer tsi/ta and lower than left dislocated
material (Topic, Focus), as illustrated in section 1.1 – while it precedes all object clitics. On the
other hand, there is considerable variation in the relative position of si and of the negation. In
Diviakë, the order is normally si nu, for instance in (21b). In Libofshë we also find a preference for
the order si nu, for instance in (22d), though the examples in (25) show that the alternative order nu
si is allowed in the same context. In Fier, if there is a single copy of si, it yields the order si nu. The
particle however can also be doubled, in which case si s nu and s nu s are equally possible orders, as
in (26b-c). The doubling of s is also systematic in Këllez, yielding s nu s as in (27). Finally, though
rare, the doubling of si in the absence of the negation is also attested, for instance in Fier in (26d).
(25)
a.
b.
(26)
a.
b.
c.
d.
ɲ
ar
dzɛs tsi
to.me they.have
said that
‘They told me not to sleep’
ma
tsə
dzɛk atsea tsi
Progr to.you I.say to.you that
‘I am telling you not to drink it’
mini mənduɛsk
tsi
s
I
think
that Prt
‘I think I don’t see him’
mini vɔi
tsi tini
si
I
want
that you Prt
‘I want you not to sleep’
tsə
dzəʃ tsi
s
nu
to.you I.said that Prt
not
‘I told you not to do it’
eu
agərʃɛ (tsi) si
z
s
Prt
nu/
not/
nu s/ s
not Prt/Prt
nu
not
dɔrm
I.sleep
L
nu o bei
not it you.drink
nu
not
u
him
vɛd
I.see
s
Prt
nu
not
dɔrɲ
sleep
s
Prt
o
it
fats
you.do
ɣa
s
Prt
kəmɛʃa
F
(27)
he
forgot that Prt
Prt
‘He forgot to wash the shirt’
ar
fudzitə
ta
s
they.have
gone
that Prt
‘They have gone out not to eat’
he.wash
the.shirt
nu
not
məkə
they.eat
(s)
Prt
K
Current accounts of Prt consider macro data such as the position of Prt with respect to the
finite complementizer and Topic/Focus material, but are not finely articulated enough to account for
subtler variation such as (25)-(27). An early attempt at discussing and capturing microvariation is
Roberts and Roussou’s (2003), who compare the evidence on Greek to that on Calabrian. For
Calabrian, they propose that the Prt mu/mi occupies the MoodP position (Rizzi’s Fin), where it is
preceded by Neg and by C/Op (Rizzi’s Force), as in (28b). On the contrary, the Greek particle
moves from Mood to C/Op, as in (28a), and this accounts for its complementary distribution with
the complementizer oti and presumably for the fact that it precedes Neg.
(28)
a.
b.
[C/OpP oti/na
[C/OpP chi/pe
[NegP no
[MP tna
[MP mu/mi
[TP . . .
[TP . . .
On the basis of (28a), one may surmise that double occurrences of the Aromanian Prt si in
(25)-(27) are to be imputed to movement from Mood to C/Op, without deletion of the lower copy.
However, doubled si can be preceded by tsi/ta (i.e. the complementizer) – which means that the
highest occurrence of si is lower than Rizzi’s Force, or Roberts and Roussou’s C/Op. Furthermore,
on the basis of a single Neg position we cannot account for the fact that Neg in Aromanian can
either be sandwiched between two copies of si or follow both of them; more than one Neg position
must be involved. In fact, variation similar to that of Aromanian characterizes Calabrian and
Sicilian varieties; the examples in (29)-(30) show the basic contrast between the Neg-Prt order in
(29) and the Prt-Neg order in (30).
(29)
a.
b.
(30)
a.
b.
vOnu
(pE) nO
mu
u hattsu
they.want
for
not
Prt
it I.do
‘They don’t want me to do it’
vu6ŒŒu pE
nnO mmu mi cama
I. want for
not
Prt
me he.calls
‘I don’t want him to call me’
tE Diku
mu
um
bi@ni
you I.tell
Prt
not
you.come
‘I am telling you not to come’
tS
aju Dittu
mu
un
mE camanu
them I.have told
Prt
not
me they.call
‘I told them not to call me’
Gizzeria
S.Pietro a Maida
Conflenti
Platania
Next, in some varieties, Prt is systematically doubled on either side of the negation, as in
(31). Other instances of doubling crop up when there are Topics or Foci, for instance in (32) –
independently of whether the basic order is Prt-Neg or Neg-Prt.
(31)
a.
b.
vOŒŒu pE
mmi nO
I. want for
Prt
not
‘I don’t want him to come’
tSi Dissi
pE
mmi
him I.told
for
Prt
‘I told him not to call me’
mi
Prt
vEni
he.comes
nO
not
mi
Prt
mi cama
me he.calls
Seminara
(32)
a.
b.
vOLLu
nO
mu
tu
mu
I.want
not
Prt
you not
‘I don’t want you to come’
suJJu kunti@ntu
mu
frati-tta
I.am glad
Prt
brother-yours
‘I am glad that your brother is not coming’
vEni
come
Gizzeria
(mu) um
Prt
not
bEna
come
Conflenti
According to Manzini and Savoia (2005), who present these data, doubling of Prt indicates
the availability of two different positions for it. The lower position is a C position, understood as we
do here, namely as a verb-related position. The higher Prt is a complementizer, understood again as
we do here, i.e. a Q(P) operator; they account for its co-occurrence with other complementizers by
invoking the iterability of complementizers (comparable to that of wh- operators). In addition, Neg
takes at least three different positions – one superordinate to the higher mu/mi, as in (32a), one
intermediate between the two copies of mu/mi, as in (31b), and one lower than the lowest copy of
mu/mi, as in (32b). In other words, they propose the ordering in (33a) – on which one must further
impose the condition (33b), that only one copy of Neg is externalized at any given time. 2
(33)
a.
b.
[Neg nO [QP mu [Neg nO [C mu [Neg nO
Condition:
only one nO externalized
Apart from the fact that (33) provides a mere encoding of the data rather than an explanation
for them, we note that it is unclear why the same element mu would turn up once with the categorial
signature C of a verb-related, modal category and once as a complementizer, which in their terms
(as in ours) is an operator over propositional content. Nevertheless, the data observed are indeed
strictly comparable to those of Aromanian. Aromanian is only slightly more restrictive than
Calabrian, in disallowing the sequencing of negation before a doubled Prt, as schematized in (34).
(34)
a.
b.
[ si
[Neg nu [ si
[Neg nu
Condition:
only one nu externalized
To a lesser extent, our empirical conclusions are confirmed by Albanian and Arbëresh
(Italo-Albanian) varieties, analysed by Manzini and Savoia (2007). In Albanian the subjunctive Prt
is tə (standard spelling të), which combines with the negation mɔs, specialized for modal
environments (imperatives, subjunctives, etc.). In the mainland variety of Gjirokastër, the orders
mɔs tə and tə mɔs are both attested, as in (35). In Arbëresh, the order is mɔs tə for instance in Vena
di Maida in (36a) but tə mɔs in Ginestra in (37a). In Vena di Maida, furthermore, mɔs/mɔhə can be
in complementary distribution with tə, as in (36b). Finally the particle tə can appear both to the right
and to the left of mɔs, as in (37b). In the variety of Barile in (38), the doubling of Prt does not
depend on the presence of the negation.
(35)
a.
b.
(36)
2
a.
tə
mɔs ɛ
Prt
not
him/her
‘Let us not dress him/her’
mɔs t
a
not
Prt
him/her
‘Let us not dress him/her’
ɐ
mə
mirə mɔs
vɛʃim
we.dress
vɛʃim
we.dress
Gjirokastër
t
ɛ
ʃɔrə
Furthermore copies of mu/mi are never adjacent. It seems reasonable to see this latter constraint as an instantiation of a
more general morphological haplology constraint (Neelemann and van de Koot 2006, van Riemsdjik 2008, Yip 1998) –
hence irrelevant for the present discussion.
(37)
(38)
it.is more good not
Prt
‘It is better that he doesn’t call him’
b.
ɐ
mə
mirə mɔhə ɛ
it.is more good not
him
‘It is better that he doesn’t call him’
a.
tə
θɛjə tə
mɔs a
to.you I.said Prt
not
him
‘I told you not to call him’
b.
iʃ
mə
mir
tə
mɔs
it.is more good Prt
not
‘It is better for you not to do it’
jamən kundɛnt
tə
(mɔs) t
I.am happy
Prt
not
Prt
I am glad not to see him’
him
he.see
ʃɔrə
he.see
Vena di Maida
θratʃə
you.call
t
Prt
a
it
a
him
ʃɔ
I.see
bətʃ
you.do
Ginestra
Barile
For independent reasons, Manzini and Savoia (2007) conclude that the negation is
specialized for the C domain, as suggested by the label NegC. This imposes rather more restrictions
on the position of the negation than in Romance, so that when there are two copies of Prt it is
constrained to appear between them, as in (38). The string in (38) also accounts for the fact that
occurrences of a single Prt yield either the sequence Prt-NegC (when the higher occurrence of Prt is
chosen) or else NegC-Prt.
(38)
[ tə
[NegC mɔs
[ tə
Despite the simplification introduced by (38) with respect to (33), the basic fact remains that
at least two positions are available to Prt. This makes a match to theories of the left periphery such
as Rizzi’s (1997), Roberts and Roussou’s (2003) not immediately obvious, nor do analyses of
complementizers as wh- pronouns help in this respect (Manzini and Savoia 2005). Needless to say,
we may enrich the models, to make them compatible with the facts; for instance we may internally
articulate Fin into several positions as proposed by Hill (2013a, b) in relation to infinitival
complements (see section 3). Nevertheless, some evidence concerning specifically Albanian tə
suggest to Manzini and Savoia (2007) that Prts may have been wrongly associated with categories
such as Mood or C – and should be reanalysed as pronominal-like. We turn to this evidence next,
which involves considering the crucial interpretive property of Prt sentences, namely that of
allowing control.
2.2
Prts and the theory of control
There is more than one reason to conclude that control is involved in Prt sentences in Aromanian, as
well as in other Romance languages (e.g. Calabrian, Romanian) and Balkan languages (Albanian).
To begin with, several examples including the si particle in section 1.1. and 2.1, have matrix verbs
that require coreference between their subject and the embedded subject. Relevant examples
involve aspectual verbs, e.g. ‘begin’ in Libofshë’s (3c) and the motion verbs ‘come’ and ‘go’ in an
aspectual interpretation in Këllez’s (9c) and (24d), or in Diviakë’s (21d). At least one example,
namely Fier’s (23d) involves a modal, namely ‘can’. Despite the fact that these verbs embed finite
complements, a disjoint reference reading of the embedded subject is not possible, exactly as it isn’t
in the infinitival sentences that they select in French/Italian/Spanish or in English. In other words,
whatever syntactico-semantic mechanism is involved in Obligatory Control, it applies irrespective
of the finite morphosyntax of the complement sentence.
A particular point of interest of Aromanian is that it preserves the common Romance
infinitive in –re (-ri in Aromanian), introduced by the preposition di/ti (cf. French/Spanish de,
Italian di). In several of the Obligatory Control environments (complements of ‘begin’,
complements of motion verbs), informants offered both a finite and an infinitival rendering, as
illustrated in (39), (40), (41). The same is true for environments that in principle allow both control
and disjoint reference, for instance verbs of command, as in (41). In (41), the infinitival
complement of Aromanian, exactly like the infinitival complements of French/Italian/Spanish or
English, is compatible only with the control construal, part of which is the distinctive de se (or here
de te, Landau 2015) reading. Evidently, the finite Prt constructs that are systematically offered as an
alternative by our informants, are interpreted in the same way.
(39)
(40)
(41)
(42)
am
agrəʃətə
tsi
s
mək/ ti
məkari
I.have forgotten
that Prt
I.eat/ for
to.eat
‘I forgot to eat’
a.
o cisiə
tsi
s
o
fak
mini/ di
fətsɛri
it I.started
that Prt
it
do
I/
for
to.do
‘I began to do it’
b.
ɲi
nɛsi
minta
tsi nu s
o fak/ di
nu
to.me came the.mind
that not Prt it I.do/ for
not
‘I remembered not to do it’
c.
am
vənit tsi
s
ti vəd/
di vədɛri (tini)
I.have come that Prt
you I.see/
for to.see you
‘I came to see you’
tsə
dzəʃ
si
s
o
fats /di
fətsɛri aist
to.you I.say Prt
Prt
it
you.do/ for to.do this
‘I tell you to do this’
a.
am
buʁitə
ta
s
u
vɛd
I.have started
that Prt
it
I.see
‘I began to see it’
a’.
mini mbu"ʁi di
gʁɛʁin
I
started for
to.speak
‘I began to speak’
b.
ei
ar
vinitə ta
s
mi vjadɛ/
ti
vidɛʁi
they have come that Prt
me they.see/ for
to.see
‘They came to see me’
D
L
fatsɛri
to.do
F
K
mini
me
Generative discussion about the nature of control spans several decades. In the early 80s,
Williams (1981) proposes that control involves the embedding of a predicative expression (not a
proposition), which is closed by the matrix subject. At the same time, Chomsky (1981)
characterizes the subject of control sentences as an empty category PRO, which has a binding
relation to some matrix argument. Within the minimalist framework, PRO is championed in
particular by Landau (2004, 2006, 2013), who proposes that control depends on Agree between a
matrix functional category (I for subject control, v for object control), which acts as the probe, and
an embedded PRO, which acts as the goal. The distribution of PRO is governed by an algorithm,
whereby only finite tensed clauses (+Agr, +T) are incapable of being controlled. The presence of a
–Agr clause or of a –Tns one licences PRO, hence control. Infinitival sentences are a typical
example of –Agr control, while finite complements in Balkan languages allow control in virtue of
their –Tns properties.
The main alternative to PRO within the minimalist framework is the movement theory of
control proposed by Hornstein (1999). Boeckx and Hornstein (2006) also consider control into
finite sentences. Normally, movement (hence control) is disallowed out of embedded sentences
because of the presence of a C phase, blocking it. However defective C phases do allow A-
movement. In order to get control into finite complements it is sufficient for them to be defective
phases. As for which sentential complements are and are not phases, Boeckx and Hornstein adopt
Landau’s idea that tenselessness is crucially involved. Finally, Manzini and Roussou (2000) are
normally considered within the bracket of movement theories of control, because they propose that
the same structure underlies both control and raising, but theirs is in reality an attempt at reviving
predication theories of control, in which control amounts to the closure of an open variable in the
embedded complement (a predicate rather than a sentence). Manzini and Roussou however have
nothing to say on control into finite sentences.
Here we are not interested directly in control, but rather in sentential introducers, hence in
subjunctive Prts. In the approaches of Landau (2004, 2006) or Boeckx and Hornstein (2006), the
contribution made by the so-called subjunctive particle to control syntax is not made explicit. Given
the requirements of Landau’s theory, or of the movement theory of control, one could propose that
the so-called subjunctive particle in reality is a –Tns morpheme – though it can equally introduce
+Tns (non-control) contexts. Yet if the particle really only contributes Tns/Mood properties, why
wouldn’t Romance languages like Italian or French or Spanish endowed with inflectional
subjunctives also licence control into their subjunctive sentences?
Manzini and Savoia (2007), Manzini (2009) seek both to develop Manzini and Roussou’s
ideas about control as a variable saturation operation (essentially a predication) – and to reinterpret
the Albanian particle të as being connected to the expression of the EPP as such a variable. A
fundamental piece of evidence tying particle të to the pronominal system is its occurrence in Linker
contexts (Campos 2008, Campos and Stavrou 2005, Franco et al. 2015). As a Lkr, të appears in
front of adjectives both as adnominal modifiers in (43a), and in predicative position in (43b). It
equally appears in front of genitives, as in (44), whether as adnominal modifiers or as predicates.
There is also a type of relative clause that appears to involve të as a Lkr, as exemplified with
standard Albanian in (45) (cf. Kallulli 2008). 3
(43)
a.
b.
(44)
a.
b.
(45)
a.
b.
ɛrðə kriatura-tə
tə
mbiðɛɲ-a
came boys-nom.def Lkr
big-pl
‘The big boys came’
jan
tə
traʃ-a
they.are
Lkr
fat-pl
‘They are fat’
ɲə/ ajɔ
kɐmb tə
matʃə-sə
a/that
leg
Lkr
cat-gen
‘a/ that leg of the cat’
kjɔ
ɐʃt
tə
ɲəriu-tə
this
is
Lkr
man-gen
‘This is of the man’s’
burri të
cilit i
ke
dhënë librin
the.man Lkr to.which to.him you.have given the.book
‘the man to whom you have given the book’
?Burri
të
cilin e
thirre
the.man
Lkr
which him
you.called
‘the man whom you called’
Vena di Maida
Vena di Maida
One may want to consider the occurrences of të in (43)-(45) in adnominal
modification/predication contexts as merely homophonous with the occurrence of të as subjunctive
particle. However a strong interpretive connection can be established between Lkr and Prt
3
Incidentally, -t(ə) also plays a role in the system of definite postnominal inflections of Albanian, realizing for instance
the masculine oblique singular in (44b); we disregard this fact here which is discussed in the references quoted.
occurrence, if we consider that a stream of formal studies takes control to be an instance of
predication (Williams 1981, Chierchia 1984). Similarly, we may take it that Lkr contexts establish a
predication between the phrase introduced by the Lkr (adjective, oblique, relative clause) and some
other DP (the head of the relative, the subject of the copular sentence etc.).
Some potential difficulties with the identification of Lkr të and Prt të are morphological. As
shown by the literature, Albanian preadjectival and pregenitival Lkrs vary according to the phifeatures, Case and definiteness of the head noun. Thus tə in the nominative plural in (43) alternates
with i in the masculine nominative singular, e in the feminine nominative singular (Campos 2008).
By contrast, Prt të is invariable. Now, crosslinguistic comparison shows that Lkr element may be
inflected (such as the Albanian article) or not inflected (such as the Persian ezafe). In languages in
which the Lkr can be inflected, the extent of which it is may vary depending on the context of
insertion (e.g. pre-adjectival or pre-genitival, cf. Franco et al. 2015). In fact, in (45) the relative
clause Lkr has an invariable të shape. Therefore the lack of an inflectional paradigm associated with
Prt të does not seem to represent an unsurmountable obstacle to its unification with Lkr të.
Importantly, occurrences of të as sentential introducer are not limited to finite subjunctive
complements. Rather in Tosk Albanian, including standard Albanian, non-finite purpose clauses
and non-finite relative clauses expressed by participial structures are introduced by (për) të, as in
(46). In contexts like (46), it appears that we can keep glossing të as Prt only to the extent that we
separate the Prt category from subjunctive modality.
(46)
a.
b.
aɾðtʃ pəɾ
t
a/i
vɛʃur
I.came for
Prt
him/them
dressed
‘I came (in order) to dress him/them’
ɲə
kəmiʃ pəɾ
t
(u)
laɾə
a
shirt for
Prt
Passiv washed
‘a shirt to be washed/to wash’
Gjirokastër
In the Arbëresh dialect of Civita, the aspectual verb ‘finish’ takes a non-inflected participial
complement preceded by tə, as in (47a), though the variant where ‘finish’ embeds a finite
‘subjunctive’ verb also exists as in (47b). Continuity between (47a) and (47b) may lead us to
conclude that the same fundamental Prt element is involved; but note once again that Prt does not
appear to have anything to do with the expression of subjunctive Mood.
(47)
a.
b.
fiɾnɔva
tə
diɛvasuɾ
ʎibɾin
I.finished
Prt
read
book-the
‘I finished reading the book’
fiɾnɔva
t
ɛ
diɛvasja
I.finished
Prt
it
I.read
‘I finished reading it’
Civita
The generative literature does not agree in its entirety on the predicative status of Lkr
structures. Thus Larson and Yamakido (2008) construe Lkrs as case assigners, while Richards
(2010) argues that they are means for identity avoidance at the PF interface. Yet generative works
treating Lkr structures as predication include den Dikken and Singhapreecha (2004), Campos and
Stavrou (2005) who argue that the Lkr is a copula. In a different take on the same general idea,
Lekakou and Szendroi 2012, Franco et al (2015) argue that the Lkr is a D, as implied by its
traditional label of article, and hence plays the role of subject of the predication.
Let us consider how this characterization of the Lkr element works in the simple example in
(43a). The structure of (43a) is an in (48); the Lkr projects a DP functional layer above the AP,
providing a lexicalization for the argument slot of the predicate ‘big’. Following Higginbotham
(1985), we assume that the interpretation of the DP ‘the big children’ requires identification of the
argument slots (theta-roles) of the adjective and the noun, which are ultimately satisfied by the
Determiner of the noun, represented in (48) by the postnominal article/definite inflection. In the
schema in (48), therefore, the Lkr D is a bound variable of the higher D closing off the DP (see
Lekakou and Szendroi 2012 on Greek, Franco et al. 2015 on Albanian for slightly different
implementations).
(48)
[kriatura [D tə]]
[DP tə [AP mbiðɛɲa]]
The characterization that Manzini and Savoia (2007) propose for Prt të is in essence the
same just reviewed for Lkr të. Specifically, “të introduces the EPP argument of the embedded
sentence as a variable … the të variable can be identified with the matrix subject, mediating
therefore the apparent coreference between the matrix subject and the embedded subject as
lexicalized by the D inflection”. Manzini (2009) further elaborates this analysis in the following
terms: “it is natural to construe të as introducing not so much a variable as a lambda-operator … the
embedded sentence is turned into a predicate-abstract assigned a value by a matrix argument”.
When applied to the minimal pair in (47), this complex of ideas leads Manzini and Savoia
(2007), Manzini (2009) to propose a novel solution concerning the categorization and positioning of
të. Prt të exactly like Lkr të is categorized as D and is taken to turn the embedded sentence into a
predicate, by abstracting on the EPP argument and opening a variable corresponding to it. Since
they keep the conventional assumption that whatever Prt does it does at the CP level, the proposed
structure for (47) is as in (49). In (49) IP is a potentially closed proposition (including an EPP
argument represented by the finite verb inflection) which is reopened by the D variable/lambdaoperator.
(49)
a.
b.
fiɾnɔva
fiɾnɔva
[CP
[CP
[D tə] C
[D tə] C
[IP ɛ diɛvasja
[IP diɛvasuɾ ʎibɾin
Though this account differs both from the movement theory of control and from Landau’s
(2013) PRO algorithm, in a more recent development, Landau (2015) himself proposes a
predicational take on control that seems compatible with (49). In discussing English predicates
which obligatorily take control complements, for instance (50a), Landau argues in favour of the
structure in (50b). D is a ‘minimal pronoun’, interpreted as a variable in its Merge position – and as
an operator (a lambda abstractor) in the [Spec, Fin] position it moves to. The projection FinP is
therefore a predicate, which gets saturated by the matrix argument John. In fact, Landau also seems
to perceive the potential connection with Lkrs structures since he comments that the position he
calls Fin (following Rizzi 1997, presumably) should really be recast as den Dikken’s (2006)
Relator. Importantly for present purposes, Landau argues that predication control is not disrupted by
the presence of a finite I. The latter may very well transmit phi-features to D by Agree, but this does
not prevent predication to take place.
(50)
a.
b.
John managed to stay healthy
… managed [Fin/RelP D
[Fin/Rel
[TP D to stay healthy
In short, different lines of research converge on the conclusion that control is a predication
operation, depending on the presence of an EPP variable, introduced by a lambda-operator. What
Manzini and Savoia (2007) add to this general picture is the specific proposal that in subjunctive Prt
languages of the Balkan type, the Prt is connected to the expression of the EPP as a variable. This
raises the important question of how Prt is to be construed in contexts that are compatible with a
control reading but do not necessitate it – or in contexts which display disjoint reference between
the matrix and the embedded subject. We return to this question in the next section.
2.3
Analysis of Aromanian
As a preliminary to the discussion of Aromanian, it is important to note that Aromanian like
Albanian has linker structures (Campos 2005, Franco et al. 2015). However unlike Albanian,
Aromanian lexically differentiates the subjunctive Prt, namely si, from pre-adjectival and pregenitival Lkrs. In fact the pre-adjectival Lkr, atseu etc. as in (51), differs from the pre-genitival
linker, which takes the form o/ali, also occurring in front of datives, as in (52).4
(51)
(52)
fitʃor-u
(a)tse-u
mar-u
boy-msg
Lkr-msg
big-msg
‘the big boy’
a.
libr-a
o fitʃor-u/
ali fet-i
book-the
Lkr boy-msg/ Lkr girl-obl.fsg
‘the book of the boy/girl’
b.
i
o am datə
o fitʃor-u/
ali
fet-i
him it I.have given
Lkr boy-msg/ Lkr girl-obl.fsg
‘I gave it to the boy/girl’
In section 2.2, the evidence connecting Albanian të to Lkr të provided us with independent
grounds for uncoupling the so-called subjunctive Prts from the I-C (Tense/Mood) system of the
sentence – and connecting it to the expression of the variable structure necessary for control.
However in Aromanian (as in many other languages, including Romanian or Greek) there is no
lexical overlapping between Lkrs and Prts. In fact, nothing that we have said predicts the necessity
of this coincidence, merely its possibility. As we have just seen, Pre-adjectival Lkrs differ lexically
from pre-genitival ones in Aromanian, though their lexical coincidence is the prevalent attested
pattern crosslinguistically. Following Albanian (48), we propose structure (53a) for the
preadjectival Lkr in (51) and structure (53b) for the pre-genitival Lkr. On the basis of the structures
in (53), the preadjectival and pre-genitival Lkr can be differentiated simply in terms of the structure
they select, namely AP and DP respectively. 5
(53)
a.
b.
[fitʃor [D u]]
[libr [D a]]
[DP atseu
[DP o
[AP maru]]
[DP fitʃoru]]
Let us then consider finite control structures embedding the Prt si, for instance (23c),
reproduced in (54a) for ease of reference. If we apply point by point the analysis of Albanian (49b)
to it, we obtain the structure in (54b), where si is assigned the status of D specialized for the context
CP/IP. In (54b) the finite verb inflection has a phi-features content sufficient to establish reference
to 1st person singular, i.e. the speaker; however the variable structures created by the presence of
Prt si (here a D element) requires to be satisfied by a matrix argument. The wellformedness of the
sentence depends on the features of the matrix argument matching those independently introduced
by the embedded finite inflection.
(54)
4
a.
mini
I
filli
began
s
Prt
lədzɛsk
I.read
un libər
a book
Incidentally, the preadjectival Lkr is not present in Romanian; the pre-genitival Lkr is present but with different
characteristics. In particular, it does not precede dative complements, and it agrees not with the genitive DP but with the
head Noun (as is normal in pre-genitival Lkr structures, cf. also Albanian).
5
Other factors may play a role, such as the agreement of pre-genitival o/ali with the embedded genitive itself, cf. fn. 4.
b.
‘I began reading a book’
…
filli
[CP
[D si] C
[IP lədzɛsk un libər
We must now consider the long-delayed question of what happens with Prt complements
that can equally receive a control or a non-control reading, or only the latter. Landau (2015) offers
the generalization that such complements are selected by propositional attitude verbs, as
exemplified in (55a) for English. Even when selecting control complements, propositional attitude
verbs have a more complex structure than the control verbs considered so far. This increased
complexity is determined by the fact that the left periphery of the sentence hosts a logophoric
centre, including crucially a representation of Author (the matrix subject, to whom the propositional
content is imputed) and/or Addressee. At the same time, the inner core of the embedded sentence
has the same structure as for predication control. Control involves binding of the DAuthor variable by
the matrix Author; the DAuthor argument can then satisfy the embedded predicate. The de se reading
effectively corresponds to Author control.
(55)
a.
b.
John intends to visit Athens
… intends
[CP DAuthor
C
[Fin/RelP D
[Fin/Rel
[TP D to visit…
Landau argues that if the embedded inflection is finite, the type of control seen in (55b), i. e.
logophoric control, is impossible – contrary to what happens in predication control. For control to
go through, the embedded control structure should be predicated of DAuthor and the latter bound by
the matrix subject – but this is impossible because DAuthor is already assigned a value by the
embedded finite I. Yet, as anticipated in presenting the Aromanian data, the syntactic and
interpretive evidence favours the conclusion that control extends to finite complements of attitude
verbs – which are in free alternation with infinitival clauses and, as far as we can tell, display the
same de se interpretation. Therefore, we keep to the fairly standard assumption that control
alternates with lack thereof in Prt complements of attitude verbs.
Let us pursue the idea that control complements of attitude verbs, for instance (23a),
repeated as (56a) for ease of reference, involve the same structure as we have already seen in (54)
above, for control complements of non-attitude verbs. We would like to suggest that predication
control is sufficient to establish the de se reading, when combined with the intrinsic properties of
the subject of attitude verbs, i.e. of representing the holder of the embedded propositional content.
(56)
a.
b.
mini vrɛm
s
I
wanted
Prt
‘I wanted to see him’
… vrɛm
[CP [D s]
o
him
vədɛm
I.saw
C
[IP o vədɛm
The question now is how non-control readings are equally allowed for Prt complements of
attitude verbs, as for instance in (24a), reproduced below in (57a); by contrast, any attempt at
introducing a non-control reading would result in ungrammaticality with non-attitude main verbs,
for instance (54). In both examples, the finite embedded verb inflection is capable of transmitting its
full set of phi-features (person, number) to the subject of predication – yet only in (57) does this
succeed in establishing reference independently of a matrix argument. Taking Landau’s suggestion,
we may assume that this has to do with the presence of a referential anchoring centre in the
embedded sentence, which we provisionally represent via the category Author in (57b). In the
picture we are offering therefore predication control and establishment of a logophoric center are
two alternative ways for satisfying the requirements of attitude verbs. In any event, we should stress
that we are offering this proposal simply for the sake of completeness. Exactly which properties of
the matrix verb (and/or which selected properties of the embedded complement) determine the
possibility of disjoint reference in (57) but its impossibility in (54) is largely outside the purview of
the present study, which concerns the contribution make to structural and interpretive properties by
sentential introducers of various kinds (complementizers, Prts, etc.).
(57)
a.
b.
vrɛu
s
u
I.wanted
Prt
it
‘I wanted him to eat it’
… vrɛu
[DAuthor
məkə
he.eats
[CP [D s]
C
[IP u məkə
From a structural point of view it is important to note that if there is a lexical subject is
realized in the embedded sentence, it takes either a left peripheral position, as for instance in (3a)
reproduced below in (58a), or a right peripheral position as in (4c), reproduced in (58b). These
examples are reproduced in (58) for ease of reference. On the contrary it is impossible for the
subject to be positioned between the Prt and the verb. This is hard to explain if the subjunctive Prt is
construed as a manifestation of Fin or Mood. On the other hand, it follows under the assumption
that the subjunctive Prt introduces an EPP variable, since this excludes that the Spec, IP position
could have lexical content. Incidentally, the generalization concerning the position of the lexical
subject holds not only in Aromanian, but also in the other languages discussed here (Albanian,
Calabrian).
(58)
a.
b.
vɔi
tsi
tini
z
iɲ
məni
I.want that you Prt
come tomorrow
‘I want you to come tomorrow’
eu
vini dənint tsi
s
vənɛm
he
came before that Prt
came
‘He came before I came’
mini
I
In the light of the discussion of control, we may also come back to structures involving the
doubling of Prt, as schematized for instance in (34) for Aromanian. The structures in (54), (56)-(57)
commit us to identifying at least one of the positions of the subjunctive Prt with the D position
available for predication/control. It is natural to assume that the other Prt position is also a D
position. Consider for instance example (26d), repeated below as (59a). If the higher copy of si
instantiates D position within the C domain, the lower copy may simply represent the IP-internal
position of the EPP argument, as in (59b).
(59)
a.
b.
eu
agərʃɛ (tsi) si
z
ɣa
he
forgot that Prt
Prt
he.wash
‘He forgot to wash the shirt’
… agərʃɛ
[CP [D si] C
[IP [D z] ɣa kəmɛʃa
kəmɛʃa
the.shirt
In other words, under the account in (59b), the doubling of subjunctive Prts is to be equated
to the doubling of subject clitics found in modal environments (e.g. questions), where both the IPinternal clitic positions and the CP-internal clitic positions are realized (see Manzini and Savoia
2005 for examples involving Northern Italian varieties). Under these conditions the negation clitic
can also vary between an IP-internal lexicalization and a CP-internal one, as schematized in (60a)
for example (26b) and in (60b) for example (26c).
(60)
a.
b.
…
…
[CP [D si]
[CP [D s] [neg nu]
C
C
[IP [D s] [neg nu]
[IP [D s]
…
…
Recall finally that in section 1 we analysed the tsi complementizer of Aromanian as a whpronoun, compatible both with a sentential restriction and an individual one. At the same time, we
left the matter open of the form that seems to substitute for tsi in the variety of Këllez, namely ta.
There are no obvious counterparts for it in other Romanian/Romance varieties – but an obvious
match for it in Albanian, namely the të particle. The borrowing of të into Aromanian leads to a
redefinition of this element, which no longer represents the subjunctive Prt, hence in present terms a
D operator/variable, but rather a complementizer, namely an operator introducing a propositional
variable. Vice versa, the subjunctive Prts si of Aromanian and să of Romanian are connected by
external evidence to the Latin and Romance si/se ‘if’ element, i.e. the polarity complementizer
(Adger and Quer 2001). In this instance, historical change causes a complementizer, i.e. an operator
introducing a propositional variable, to shift to the role of Prt, i.e. in present terms a D element.6
2.4
Conclusions
In section 1, we argued that the so-called Romance complementizer, or CForce in the split CP
framework of Rizzi (1997), is in reality a wh-pronoun which turns the propositional content into a
free relative. There is no sense therefore in which the sentential introducer is the highest functional
projection in the sentence. The highest functional projection is the highest position of V, for which
we keep the conventional label C (the phase head). In this section, we have seen that subjunctive
Prts, which have been argued to represent lower complementizer positions, i.e. CFin in Rizzi’s
(1997) schema, are better reanalysed as linker (D) elements connected to the lexicalization of the
subject of the sentence. Therefore, the overall picture of sentential organization that we propose is
closer to the IP-CP structure of Chomsky (1995) than to more articulated cartographic proposals.
Specifically, the microvariation evidence we consider does not favour a refinement of functional
hierarchies (e.g. two Fin positions for the doubling of Prts) over more conservative solutions.
3.
Non-finite complements and Prepositional introducers
3.1
The evidence
In standard Romanian there are two types of non-finite complement sentences, which involve either
the so-called short infinitive or the supine. Morphologically, the short infinitive corresponds to the
pure verb base, i.e. to the root followed by thematic vowel, while the ‘long’ infinitive in –re has
become a de-verbal Noun. The short infinitive can be bare, or it can be preceded by the sentential
introducer a. Interestingly, Early Romanian had a larger repertory of infinitival structures, since it
also admitted of long, i.e. –re inflected, infinitives. Furthermore the de Preposition could precede
a+infinitive sequences. In recent literature influenced by Rizzi’s (1997) model of the left periphery,
the elements a and de are identified with Fin; where the two co-occur it is assumed that the Fin
position splits into two, with the higher Fin taken by de and the lower Fin taken by a (Hill 2013b).
The second type of non-finite sentential complement in standard Romanian involves the socalled supine. Morphologically, the supine is the perfect participle form, which takes the label
supine in that it has an active construal rather than a passive/unaccusative one. In Early Romanian,
according to Hill (2013b) “the supine starts in non-finite relatives, expands to predicative BE
structures, and displays timid distribution as sentential complement to verbs by the end of the 18th
century”. Relatives, copular contexts and complements to certain verbs are the three main context
6
Something very similar happens with Ancient Greek hina vs. Modern Greek na (Roberts and Roussou 2003). The
place of the polarity complementizer in Aromanian hypotheticals appears to be taken by ma, which is further attested as
progressive particle. The English equivalent would be if one expressed hypotheticals by gerunds, e.g. ‘you coming, I
will wait for you’. In standard Romanian are introduced by a complex form dacă.
for the supine in standard Romanian as well. All three are introduced by the preposition de, which
for Hill (2013a), again occupies the Fin position.
The Aromanian evidence matches Early Romanian more closely than the standard in several
respects, for instance in the use of infinitives presenting the pan-Romance infinitival inflection –r(e)
(so-called long infinitives). In externalist terms, this is particularly notable, in that Aromanian is
spoken in contact with languages (Tosk Albanian) that have fully undergone the shift to infinitivalless status. At the same time, Aromanian therefore is far from atypical in the Romance panorama.
For instance, the Calabrian varieties with mu/mi Prt complements, preserve the morphological
infinitive –re, typically under modals, and sometimes extended to other obligatory control contexts..
In Aromanian the long infinitive occurs in many of the contexts which present it in more
familiar Romance languages, including complement control clauses of the kind exemplified in (39)(42) above, where the infinitive alternates with the subjunctive Prt construction. All examples of
infinitives display control, either by an antecedent, or so-called arbitrary control (i.e. generic,
quantificational reference of the control variable), except for causative embeddings, for instance
(62e), (63d). Control environments include complements of aspectual, modal and attitude verbs, as
in (39)-(42), (62a-c) and (63a-b), as well as infinitival relatives. In the latter context, the infinitive
alternates with the subjunctive Prt construction, as in (61b), (64a). In all instances, the long
infinitive is preceded by the Preposition ti/di, obviously related to pan-Romance genitive
Preposition di/de ‘of’; in Aromanian, where genitive is a case, the basic meaning of ti/di appears to
be close to English ‘for’, as in (61a).
(61)
a.
b.
c.
d.
(62)
a.
b.
c.
d.
e.
(63)
a.
esti
libra
ti
tʃələməɲ
is
the.book
for
children
‘It is the book for children’
esti pənə
ti məkari/tsi s
mə"kəm
is bread
for to.eat/ that Prt
we.eat
‘there is some bread to eat’
aist jinu
esti
ti
vəndɛri
that wine
is
for
to.sell
‘That wine is for selling’
vin
ti
vədɛri (atseu)
I.came for
to.see (him)
‘I came to see him’
mbari
ti
məkari
I.stopped
for
to.eat
‘I stopped eating’
ɲ
ɛrəseʃti
di/ ti vədɛri
to.me it.likes
for
to.see
‘I like seeing him’
ma
tsə
dzɛk di
fətsɛri
Progr to.you I.say for
to.do
‘I am telling you to do it’
esti
unə
kəmiʃa
di
ɣari
is
a
shirt
for
to.wash
‘It is a shirt to be washed’
i-o
fɛtʃ
kəmiʃa
otsui
to.him-it
I.made
the.shirt
to.that.one
‘I made him wash the shirt’
mini viɲ
di
vədɛri (tini)
I
came for
to.see you
‘I came to see you’
D
L
di
for
ɣari
to.wash
F
b.
c.
d.
(64)
a.
b.
eu
mbari di
(u) lədzɛri libru
he
stopped for it to.read the.book
‘He finished reading the book’
aist kəmɛʃ
esti
di
ɣari
this shirt
is
for
to.wash
‘This shirt is to be washed’
mini u
fak
d
ardɛri
I
him make
for
to.laugh
‘I made him laugh’
aistə esti
jin
ti
(ni)
bɛʁi/ ta
s nu s
this
is
wine for
not
to.drink/ that Prt not Prt
‘This is wine not to be drunk’
aistə libʁə esti
ti
(ni)
kəntaʁi
this
book is
for
not
to.read
‘This book is not to be read’
u bɛm
it we.drink
K
The other major non-finite construction of Aromanian employs the perfect participle with
active interpretation, i.e. what in Romanian linguistics is known as the supine. This has a much
more limited distribution than in standard Romanian, since it is limited to adjuncts introduced by ni
‘without’. Note that control is not obligatory, as one may expect to be the case in adjuncts, cf. (66a)
and Mensching (2000) for cross-Romance comparison. The embedded verbs in (65)-(67) all have
the –t(ə) morphology which characterizes the Latin and Romance perfect participle.
(65)
a.
b.
(66)
a.
b.
(67)
a.
b.
(68)
a.
b.
mini am
inʃətə
ni
məkatə
I
have gone.out
without
eaten
‘I went out without eating’
atseu fudzi ni
vədzutə
(pi mini)
he
went without seen
(to me)
‘He left without seeing me’
am
fudzit ni
vənit tini
I.have gone not
come you
‘I left without you coming’
ɲisai
ni
vədzut
tini
I.went.out
without
seen
you
‘I went out without seeing you’
mini iʃɛjə
ni
ti
vidzut (tini)
I
went.out
without you seen (you)
‘I went out without seeing you’
mini fudzi ni
məkat (pəni)
I
went without eaten (the.bread)
‘I left without eating the bread’
am
iʃatə
ni
durɲit / məkatə pɛna
I.have gone.out
not
slept / eaten the.bread
‘I went out without sleeping/eting the bread’
ei
fudzire
ni
gʁitə a ɲia
they went
without
called to me
‘They left without calling me’
D
L
F
K
A close comparison can be run between the Aromanian supine and Tosk Albanian (the
language in contact), where the preposition pa ‘without’ also introduces a perfect participle
(inflected by –r in Albanian), as in (69).
(69)
dɔla
pa
ɛ
lar
I.went.out
without
it
wash
‘I went out without without washing it’
Gjirokastër
The data of Geg Albanian (varieties of Northern Albania and the Kosovo) are also worth
reporting in this connection. Geg Albanian differs from Tosk Albanian in that perfect participles,
embedded under ‘have’ or ‘be’ in the perfect tenses, do not present an –r inflection, but rather
consist of bare verb bases, as in (70). The same form of the verb shows up under the preposition mɛ
‘with’, in what is traditionally known as paskajore ‘infinitive’, in Albanian linguistics. In essence,
the paskajore is found under obligatory control and attitude verbs, allowing both control readings,
as in (71a-b), and non-control readings, as in (71c). Unsurprisingly, the paskajore can be embedded
under the preposition pɑ ‘without’, yielding the equivalent of Tosk Albanian (69a), as in (71d).
(70)
a.
b.
(71)
a.
b.
c.
d.
ɛ
kan
mlu
him they.have
covered
‘They have covered him’
ɛ
kam lɒ
him I.have washed
‘I have washed him’
i
kam θɒ̃:n m
u
mlu
to.him I.have said to
MP cover
‘I told him to cover himself’
du
mɛ
ɛ
lɒ
I.want to
him/her
wash
‘I want to wash him/her’
doin
(tʃi) mɛ
ɒ:ɾð ata
they.want
that to
come they
‘They want to come’
dola
pɒ
u
la:
I.went.out
without
MP wash
‘I went out without washing myself’
Shkodër
In short, the varied repertory of non-finite embedding that we have now reviewed involve
participial forms, infinitival forms and bare verb bases; despite the differences they share the
property of being introduced by Prepositions. These include in particular di/ti ‘of’ in front of the
long infinitive in Aromanian, and ni ‘without’ in front of the perfect participle (supine).
3.2
Prepositions as sentential introducers
Rizzi (1997) for Italian proposes that Prepositional introducers should be identified with
complementizers, and specifically CFin. A similar proposal is elaborated by Hill (2013) for
Romanian, who argues that Fin should be split into two positions to account for sequences de a in
Early Romanian. This solution is natural within a cartographic account of the left periphery, but
once we reduce finite complementizers to wh- pronouns (section 1) and subjunctive Prts to D
linker-like elements (section 2), it would be strange if we had to revert to C for the sentential
introducers that most obviously belong to a different category than C, namely P. If on simplicity
grounds we reject the conclusion that elements like Aromanian ti/di bear the double categorization
P and C (Fin), then we are faced with the question why certain Prepositions can and must embed
non-finite complements.
In order to answer this question, we must go back to the proposal of section 1, that so-called
complementizers are in reality wh- pronouns. What we have failed to address is the question why.
Why couldn’t propositional content be embedded without the benefit of a wh- operator? In this
connection, it is worth remembering that typologically, it is normal for embedded sentences to take
a format quite radically different from that of main sentences. One strategy that is particularly
widespread and well-studied involves the nominalization of the sentence. While the verbal form
takes on a nominal inflection, the case array is rearranged, so that in particular the subject turns up
as an oblique (genitive). This kind of complementation is well-known from Indo-Aryan or from
Turkish (Kornfilt 1997). Similarly, in the perspective of section 1, the embedding of a finite
sentence in Romance is a form of nominalization. This is particularly evident in approaches such as
Kayne (2010), Arsenjievic (2009) where finite embedded sentences are relative clauses headed by
an abstract noun; the same is true of Manzini and Savoia’s (2011) approach where the embedded
complement is effectively a free relative.
The question then arises why nominalization would be necessary for propositional
embedding. The standard minimalist theory of argument embedding provides a fairly intuitive
answer to this question – which, as we will see, also yields an insight into why Prepositions should
be found as sentential introducers. Embedding of a DP in one of the core argument positions of the
sentence involves an Agree operation; in an ordinary transitive sentence, the internal argument
Agrees with v, licencing accusative Case– the external argument agrees with I licencing nominative
case. Now, suppose that instead of DP we tried to embed a CP. A DP has a phi-feature set that can
satisfy the v or I probe – but CP is radically deprived of such features. Therefore it cannot enter into
an Agree relation, which on the one hand leaves the probe without a suitable goal – on the hand,
deprives the sentential complement of case and hence of Visibility. From this point of view, there is
no Case Resistance of CPs in the sense of Stowell (1981) – rather a resistance to Agree, as in (72).
(72)
Agree Resistance Theorem
CPs cannot enter into Agree relations with v, I probes because of their lack of phi-features.
The nominalizing strategies adopted by many languages in order to embed sentential
complements are a way to circumvent the impossibility of embedding CPs. The treatment of
sentential complements as (free) relatives, either headed by an abstract nominal or by a wh- element
(the so-called complementizer), is the way of nominalizing sentential content adopted by the
Romance languages.
This does not yet tell us anything about the Prepositional introducers that are obligatory in
order to embed infinitival sentences in Romance. The last decade of studies on the internal structure
of complex PPs (Svenonius 2006) has clarified the different roles of lexical elements generically
classified as Ps within such structures. The P elements that we have seen to introduce infinitival and
participial clauses in Romance and Albanian are all oblique Case markers – belonging to what is
sometimes notated as a K category. Now, note that the standard minimalist Case licencing via
Agree only applies to direct cases; oblique cases (genitive, dative, instrumental) are not covered by
this mechanism. In other words, oblique embedding does not fall under (72). Therefore we propose
that in the Romance languages (in Geg Albanian etc.) one way to get around the impossibility of
licencing sentential constituents via Agree is to turn sentences, specifically non-finite (infinitival,
participial) sentence into obliques – by introducing them with Prepositions ‘of’ (genitive), ‘to’
(dative), ‘with’ (instrumental).
Though the matter was left implicit in section 1, it appears that complementation via a whpropositional pronoun, i.e. via relativization, is restricted to finite sentences. However there is no
necessary connection between the obliquization strategy, i.e. P introducers, and non-finite
complements – it is just a preferential route (for Romance). 7 A case in point is Early Romanian, as
7
The cropping up of oblique Prepositions in front of the infinitive reminds us of the fact that external evidence connects
Indo-European infinitival morphology to the oblique forms of a verbal noun (e.g. Szemerényi 1999: 324-325).
illustrated by Hill (2013b), where the de preposition could also precede finite complements, as in
(73). As noted by Hill (2013) in Early Romanian de “heads possessives, complements of origin,
‘by’ phrases, complements of location”, establishing its bona fide P categorization (cf. DobrovieSorin et al. 2013 on standard Romanian).
(73)
au poruncitŭ de au făcut
un sicreiu
has ordered of have made a coffin
‘He has ordered them to make a coffin’ (Hill 2013b)
Sardinian varieties (Jones 1993) provide an interesting example of complement sentences
which can equally be introduced by a k- complementizer or by a preposition, namely so-called
inflected infinitives, as illustrated in (74).
(74)
l an
fattu innantis dE/ki EnnErE-„E
it they.have done before of/that come-2sg
‘They did it before you came’
Dorgali (Sardinia)
Let us then provisionally conclude that the apparent complementary distribution of k- and
Prepositional introducers in finite and non-finite sentence is the product of external factors, rather
than of internal necessity. Other problems arise for the proposal that Prepositional introducers really
belong to the category P – and more specifically that they are enlisted as oblique case assigners. For
instance, we may expect di/de or a sentential complements in Romance to be resumed in the same
way as ordinary di/de-DP or a-DP expressions. Indeed di/de-CP and di/de-DPs can be resumed by
accusative clitics, as exemplified by Italian (75) – though resumption by partitive/genitive ne is only
possible in for di/de-DPs. In (76) an a-CP complement can be resumed by a locative clitic, as an
inanimate a-DP object – though it is not clear why the latter is more easily resumed by a dative.
(75)
a.
c.
(76)
a.
b.
Ne/la troverò ,
dell’acqua
of.it/it I.will.find
of.the water
‘I will find water’
*Ne/lo penso, di essere malato
of.it/it I.think of to.be sick
‘I think I am sick’
Ce/?*glie
l’
ho
costretto, ad andarci
to.it
him I.have forced to
go-there
‘I forced him to go there’
Ci/?gli ho
dato una stirata alla camicia
to.it
I.have given an ironing to.the shirt
‘I gave an ironing to the shirt’
Another potential source of problems has to do with subject sentences. To the extent that
they can surface bare in languages like Italian (77a) we must perhaps invoke their Topic status.
Vice versa, we do expect subject sentences to be able to appear preceded by oblique case-markers
to the extent that they may be construed as quirky subjects. This is likely to be what happens in the
impersonal passive in (77b).
(77)
a.
b.
Mangiare bene è importante
to.eat well
is important
‘It is important to eat well’
*(Di) aver parlato è stato confessato
of
to.have talked has been confessed
solo da Paolo
only by P
‘Only Paolo confessed speaking’
In other words, it seems to us that there no obvious grounds for dismissing the idea that
Prepositions introducing non-finite sentences in Romance are anything other than the genitive,
dative, etc. case markers that also appear in front of DPs. In the present account, the bases for such
a construct are posed by the Agree Resistance principle in (72), which can be circumvented either
by nominalizing complement sentences – or else by rendering them as obliques. In any event, in
evaluating the proposal that Prepositional introducers do belong to the category P and specifically
contribute oblique case (genitive, dative, instrumental) to the embedded sentence, we should also
consider what the available alternatives are. As mentioned in introducing the data, the leading
alternative is that elements such de/di, a lexicalize C positions, perhaps Fin in an articulated left
periphery of the type proposed by Rizzi (1997). This is equivalent to saying that these elements
systematically belong to two categories, namely C and P. The question then is why this is so.
Asking why certain lexical elements (always case assigners) can be merged as both P and C
amounts to seeking what properties P and C may have in common. In other words, even in a
grammaticalization perspective one must eventually propose why P properties make our introducers
likely candidates for Fin. As far as we can tell, the literature treating Ps as complementizers
provides no answer.
4.
Conclusions
In what precedes, we considered in some detail the complementation system of Aromanian – which
has also led us to review similar facts of other Romance languages and of the language spoken in
contact with Aromanian, namely Albanian. As we noted at the end of section 2, the overall picture
of sentential organization that we propose is closer to the IP-CP structure of Chomsky (1995) than
to more articulated cartographic proposals. In the studies available within the split CP model of the
left periphery, the sentential introducers of Aromanian would find a natural collocation either as
exponents of CForce (finite complementizers) or as exponents of CFin (subjunctive Prts, Prepositional
introducers), eventually requiring certain refinements in the hierarchy.
We have proposed a different picture, where sentential introducers provide a response to the
difficulty of embedding sentential (CP) as opposed to nominal (DP) content in argument position.
These responses are of two main type in Romance. In section 1, we argued that the Romance kcomplementizers are wh-pronouns which turn propositional contents into free relatives, effectively
nominalizing them. This allows the relativized sentence to be embedded in ordinary object and
subject position, circumventing the Agree Resistance Theorem formulated in (72). In section 3, we
have concluded that a different strategy is employed with non-finite complements (infinitive,
participles), namely obliquization. The Prepositional introducers of Romance are not only Ps, but in
fact case markers, exactly as in their occurrences embedding a DP. Making a sentential content into
an oblique is a different way of excluding it from the consequences of the Agree Resistance
Theorem.
Finally, in section 2 we considered as set of sentential introducers that characterize
Balkan/Romance languages ‘without infinitives’, or more properly, with control into finite
(‘subjunctive’) sentences. Following Manzini and Savoia (2007) on Albanian, we proposed that in
Aromanian as well, Prts are neither a Mood projection of the verb, nor are they complementizers of
sort. Rather they should be connected to the linkers system and recognized as D(-related)
categories, introducing the EPP argument as a variable and allowing control by a matrix argument.
References
Adger, David, Quer, Joseph. 2001. The syntax and semantics of unselected embedded questions.
Language 77: 107-133.
Arsenijević, Boban. 2009. Clausal complementation as relativization. Lingua 119, 39-50.
Bayer, Josef. 2001. Two grammars in one: Sentential complements and complementizers in Bengali
and other South Asian languages. The Yearbook of South Asian Languages: Tokyo
Symposium on South Asian Languages - Contact, Convergence and Typology, P.
Bhaskarorao and K. V. Subbarao (eds.), New Delhi: Sage Publications, 11-36.
Besten, Hans den. 1984. On the interaction of root transformations and lexical deletive rules, in W.
Abraham (ed.), On the formal syntax of West-Germania, Amsterdam: Benjamins, pp. 47131.
Boeckx, Cedric, Hornstein, Norbert. 2006. The virtues of control as movement. In Davies and
Dubinsky (eds.) Special Issue on Raising and Control. Syntax 9.2
Bresnan, Joan. 1972. Theory of complementation in English syntax. Ph. D. Dissertation, MIT;
Cambridge, Mass.
Campos, Hector. 2005. Noun modification, pseudo-articles, and last resort operations in
Arvantovlaxika and in Romanian. Lingua 115: 311–347
Campos, Hector. 2008. Some notes on adjectival articles in Albanian. Lingua 119, 1009-1034.
Campos, Hector, Stavrou, Melita, 2005. Polydefinites in Greek and Aromanian. In: Tomić, O. M.
(Ed.), Balkan syntax and semantics. John Benjamins, Amsterdam, pp. 137–173.
Chierchia Gennaro. 1984. Topics in the syntax and semantics of infinitives and gerunds. Ph.D.
Dissertation, University of Massachusetts at Amherst
Chomsky, Noam. 1995. The minimalist program. MIT Press, Cambridge, MA.
Dikken, Marcel den, Singhapreecha, P., 2004. Complex Noun Phrases and linkers. Syntax 7, 1–54.
Dikken, Marcel den, 2006. Relators and linkers. MIT Press, Cambridge, MA.
Dobrovie-Sorin, Carmen. 1994. The syntax of Romanian. Walter De Gruyter, Berlin.
Dobrovie-Sorin, Carmen, Nedelcu, Isabela, Giurgea, Ion. 2013. Genitive DPs and pronominal
possessors. In: Carmen Dobrovie-Sorin and Ion Giurgea (Eds) A Reference Grammar of
Romanian: Volume 1: The noun phrase, pp. 309–354. Amsterdam: John Benjamins
Franco, Ludovico, Manzini, M. Rita, Savoia, Leonardo M. 2015. Linkers and agreement. The
Linguistic Review 32: 277 – 332
Giorgi, Alessandra. 2009. About the Speaker: Towards a Syntax of Indexicality. Oxford: Oxford
University Press
Haegeman, Liliane. 2012. The syntax of MCP: Deriving the truncation account. In Lobke
Aelbrecht, Liliane Haegeman, Rachel Nye. Main Clause Phenomena, Amsterdam: John
Benjamins
Higginbotham, James. 1985. On semantics. Linguistic Inquiry 16, 547-621.
Hill, Virginia 2013a. The emergence of the Romanian subjunctive. The Linguistic Review 30: 547 –
583
Hill, Virginia. 2013b. The emergence of the Romanian supine. Journal of Historical Linguistics 2
Hornstein, Norbert. 1999. Movement and Control. Linguistic Inquiry 30: 69-96
Jordan, Maria 2009. Loss of infinitival complementation in Romanian diachronic syntax, Ph. D.
Dissertation, University of Florida
Joseph, Brian. 1983. The synchrony and diachrony of the Balkan infinitive, Cambridge: Cambridge
University Press.
Kallulli, Dalina. 2008. Resumption, relativization, null objects and information structure. In J.
Hartmann, V. Hegedus & H. van Riemsdijk (eds.) Sounds of Silence: Empty Elements in
Syntax and Phonology 235-264. Amsterdam: Elsevier.
Kayne, Richard 1976. French relative “que”. Current Studies in Romance Linguistics, Fritz Hensey
and Marta Luján (eds.), Washington, DC: Georgetown University Press, 255–299.
Kayne, Richard. 2010. Comparisons and contrasts. Oxford University Press, New York, pp. 190227.
Kornfilt, Jaklyn.1997. Turkish. London: Routledge
Landau Idan. 2004. The scale of finiteness and the calculus of control. Natural Language and
Linguistic Theory 22: 811-87
Landau Idan. 2006. Severing the distribution of PRO from Case. In Davies and Dubinsky (eds.)
Special Issue on Raising and Control. Syntax 9.2
Landau Idan. 2013. Control in generative grammar. A research companion. Cambridge: Cambridge
University Press
Landau Idan. 2015. A two-tiered theory of control. Cambridge, Mass: The MIT Press.
Larson, Richard, Yamakido, H. 2008. Ezafe and the deep position of nominal modifiers. In:
McNally, L., Kennedy, C. (Eds.), Adjectives and adverbs. Syntax, Semantics, and Discourse.
Oxford University Press, Oxford, pp. 43-70.
Ledgeway, Adam. 2003. Il sistema completivo dei dialetti meridionali: La doppia serie di
complementatori, Rivista Italiana di Dialettologia 27: 89-147.
Ledgeway, Adam. 2005. Moving Through the Left Periphery: The Dual Complementiser System in
the Dialects of Southern Italy, Transactions of the Philological Society 103: 339-96.
Lekakou, Marika, Szendrői, Kriszta. 2012. Polydefinites in Greek: Ellipsis, close apposition and
expletive determiners. Journal of Linguistics 48, 107-149.
Manzini, M. Rita 2009. ‘Pro, pro and NP-trace (raising) are interpretations’, in K. Grohmann (ed.),
Phase theory: Features, Arguments, Interpretations. Amsterdam/Oxford: Elsevier, pp. 131180.
Manzini M. Rita, Roussou Anna. 2000. A minimalist theory of A-movement and control. Lingua
110: 409-447
Manzini, M. R., Savoia. L. M. 2003. The nature of complementizers. Rivista di Grammatica
Generativa 28, 87–110.
Manzini, M. Rita, Savoia. Leonardo M. 2005. I dialetti italiani e romanci. Morfosintassi
generativa. Alessandria: Ed. dell’Orso
Manzini, M. Rita, Savoia. Leonardo M. 2007. A Unification of Morphology and Syntax.
Investigations into Romance and Albanian dialects. Routledge, London.
Manzini, M. Rita, Savoia. Leonardo M. 2011. Grammatical Categories. Cambridge University
Press, Cambridge.
Mensching, Guido. 2000. Infinitive Constructions with specified subjects. New York: Oxford
University Press.
Neeleman, Ad, van de Koot, Hans. 2006. Syntactic Haplology. In Martin Everaert and Henk van
Riemsdjik (eds.), The Blackwell Companion to Syntax, Vol. IV. Blackwell, Oxford, pp. 684710.
Richards, Norvin. 2010. Uttering Trees. MIT Press, Cambridge, MA.
Riemsdijk, Henk van. 2008. Identity Avoidance: OCP-effects in Swiss Relatives. In: Freidin, R.,
Otero, C. P., Zubizarreta, M. L. (Eds.), Foundational Issues in Linguistic Theory. Essays in
Honor of Jean-Roger Vergnaud. MIT Press, Cambridge, MA, pp. 227-250.
Rivero, Maria-Luisa. 1994. Clause structure and V-Movement in the languages of the Balkans.
Natural Language and Linguistic Theory 12: 63-120
Rizzi, Luigi. 1990. Relativized Minimality. The MIT Press, Cambridge, Mass.
Rizzi, Luigi. 1997. The fine structure of the left periphery, in L. Haegeman (ed.), Elements of
grammar. Dordrecht: Kluwer, pp. 281-337
Roberts, Ian, Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to
Grammaticalization. Cambridge: Cambridge University Press.
Rosenbaum, Peter. 1967. The grammar of English complement constructions. Cambridge, Mass.:
The MIT Press
Santorini, Beatrice. 1989. The generalization of the Verb-Second Constraint in the History of
English. Ph. D. Dissertation, University of Pennsylvania, Philadelphia.
Svenonius, Peter. 2006. The emergence of axial parts. Nordlyd 33. 1–22.
Stowell, Timothy. 1981. Origins of phrase structure. Ph. D. Dissertation, MIT, Cambridge, Mass.
Szemerényi, Oswald. 1999. Introduction to Indo-European Linguistics. Oxford: Oxford University
Press
Williams, Edwin. 1981. Predication. Linguistic Inquiry 11: 203-238
Yip, Moira. 1998. Identity Avoidance in Phonology and Morphology. In: Lapointe S. G., Brentari
D. K., Farrell, M. (Eds.), Morphology and Its Relation to Phonology and Syntax. CSLI
Stanford, CA, pp. 216-246.